We want to install all dependencies ourselves to know exactly what we
get.
I accidentally got some wraps built when I did not expect so.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.com>
Move the ad hoc filling code into a separate function. Then we can
easily add an alternative implementation of the new function using
libdisplay-info without messing up the code any more than necessary.
Pure refactoring.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.com>
Now that this is used only internally in modes.c, move it there. It will
not be used with libdisplay-info.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.com>
This will make adding libdisplay-info as another EDID parser easier,
because libdisplay-info always returns malloc'd strings.
To make things easier to extend as well, I introduce struct
drm_head_info. The libdisplay-info case will likely return more
information than this in the future.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.com>
Global arguments leak into Meson subprojects. Let's not do that.
Specifically, -fvisibility=hidden leaks into a future sub-project
libdisplay-info, where it results the DSO not exporting any symbols.
Libdisplay-info uses a linker script to define the exported symbols and
not visiblity.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.com>
There are some ICC profiles that contain something named VCGT tag. These
are usually power curves (y = x ^ exp) that were loaded in the video
card when the ICC profile was created. So the compositor should mimic
that in order to use the profile.
Weston already has support for that, but our ICC profile tests were
missing this case. This adds such tests.
For testing purposes, we have added tests with different exponents per
color channel.
Signed-off-by: Leandro Ribeiro <leandro.ribeiro@collabora.com>
Without that we may crash when trying to create a PTYPE_CLUT ICC profile
with dimension zeroed. This would be wrong, so with this change we are
basically validating the test case arguments.
Signed-off-by: Leandro Ribeiro <leandro.ribeiro@collabora.com>
For now we have some tests with the same name. Differentiate them based
on the ICC profile type that they build: CLUT vs matrix shaper.
Signed-off-by: Leandro Ribeiro <leandro.ribeiro@collabora.com>
In Pipewire and Gstreamer terminology Weston is a "live" source (as we
do not explicitly set PW_KEY_STREAM_IS_LIVE to false).
Such sources, be it compositors, cameras or microphones, usually set
the current system time as timestamps on buffers in order to make life
easier for consumers. Thus let's do so as well.
This notably helps when recording using `gstpipewiresrc` with the
`keepalive-time` property set.
Signed-off-by: Robert Mader <robert.mader@collabora.com>
instead of the output mode. The mode doesn't say anything about the
actual output geometry which could lead to buffers extending the output
region on rotated monitors. This now also works with moving the window
to different monitors.
Signed-off-by: Sebastian Wick <sebastian.wick@redhat.com>
Due to the latest developments fdo no longer allows new users to fork
the Weston project so let's inform users about that.
This also swaps 'Finding something to work on' with 'Sending patches'
paragraph as the first thing users need to look into.
Signed-off-by: Marius Vlad <marius.vlad@collabora.com>
As of PipeWire version 0.3.69, the gstpipewiresrc element uses the
existence of a modifier as a trigger to select dmabuf memory, failing
caps negotiation as we don't send DMA buffers yet.
Remove the linear modifier for now, to be added back when we add dmabuf
support to the PipeWire backend. This allows testing the PipeWire
backend with current GStreamer + PipeWire.
Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
Pass the backend instead of the compositor to the PipeWire output API
create_head() method and increment the API version.
That way the backend will not have to find the backend pointer from the
compositor. This is trivial now, but in the multi-backend case would
entail iterating over all backends to find the correct one.
Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
The session_listener is embedded in the DRM backend structure.
Use this to obtain the DRM backend with container_of().
That way the DRM backend will not have to be found from the compositor.
This is trivial now, but in the multi-backend case would entail
iterating over all backends to find the correct one.
Signed-off-by: Philipp Zabel <philipp.zabel@gmail.com>
Now that our process-launching internals are identical between the
(still-misnamed) weston_client_launch and the frontend's Xwayland
launcher, we can reuse the internals instead of open-coding it.
As a result, we now additionally prevent Xwayland from inheriting
Weston's signal mask, by clearing SIG_UNBLOCK on all signals. This
should have no observable effect as we do not depend on signal handling
within Xwayland, instead using the displayfd readiness mechanism since
c2f4201ed2.
Signed-off-by: Daniel Stone <daniels@collabora.com>
This gets us closer to the implementation of weston_client_launch, so we
can reuse that instead of open-coding it.
Signed-off-by: Daniel Stone <daniels@collabora.com>
weston_client_start() takes only a single path with no arguments,
forking a process to start that command line, and creating a client from
it.
weston_client_launch(), which was always misnamed and will be renamed in
the next patch, now only handles the child process and nothing else.
Signed-off-by: Daniel Stone <daniels@collabora.com>
When we launch a child, we need to clear CLOEXEC on any FDs we want to
survive the exec. Use an array for doing this, so it's more generic and
we can allow callers to pass in their own.
Signed-off-by: Daniel Stone <daniels@collabora.com>
See discussion in wayland/weston!951 for the reasoning why: the
screenshooter must only deal with wl_client.
Signed-off-by: Daniel Stone <daniels@collabora.com>
When we're asked to take a screenshot but are already taking one, just
exit out of the function early.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Use assert() to check for invalid NULL arguments. Something like that happens
during development and assert() makes it easier to find the error. And it avoids
unnecessary additional error handling.
The hmi-controller asserted anyways so this just moves the assert on level
deeper. Other controller probably do the same thing, or don't check the return
values at all.
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
The return values for most of the API functions will be removed and replaced by
asserts. So remove the return value checks. The end result will be the same:
These functions only fail for incorrect API usage, so basically the asserts are
moved from the hmi-controller into the shell.
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
The return values for most of the API functions will be removed and replaced by
asserts. So checking return values will no longer work and passing invalid
arguments will trigger asserts. Modify and remove the tests accordingly.
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
There are V4L2 devices that will output NV12 but will do so using one dma
buffer. To support this, we need to add the same dma buffer twice but with
a different offset for the chrominance plane.
Also supports situations of 3 planes (e.g. YU12) inside a single dma buffer.
Fixes: #712
Signed-off-by: Bram Stolk (b.stolk@gmail.com)
Add documentation for the PipeWire backend.
Co-authored-by: Michael Tretter <m.tretter@pengutronix.de>
Signed-off-by: Philipp Zabel <philipp.zabel@gmail.com>
Signed-off-by: Michael Tretter <m.tretter@pengutronix.de>
Add a separate PipeWire backend based on the PipeWire plugin. The backend
requires PipeWire 0.3.x.
The PipeWire backend can be used as a standalone-backend backend for streaming
and composing Wayland clients to PipeWire.
The backend supports the on-demand creation of heads via the
weston_pipewire_output_api_v1. It also supports per-output pixel format
configuration via a gbm-format option.
Multiple PipeWire outputs can be created by setting the num-outputs option in
the [pipewire] section.
Co-authored-by: Michael Tretter <m.tretter@pengutronix.de>
Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
Signed-off-by: Michael Tretter <m.tretter@pengutronix.de>
Currently, if a head is detached, the entire state of the device is invalidated
to make sure that the connector is disabled on the next atomic commit. Side
effect of the invalid state is that all planes are disabled on the next commit.
This includes planes that are used with a different head that is not part of the
next atomic commit. Disabling the planes of unrelated outputs causes a blanking
of these outputs until output is repainted and the plane is reenabled.
Store the detached heads in a list on the output and disable the connectors for
all heads in this list in the next atomic commit.
Signed-off-by: Michael Tretter <m.tretter@pengutronix.de>
Pass the VNC backend to vnc_head_create().
That way the already known backend will not have to be found from the
compositor. This is trivial now, but in the multi-backend case would
entail iterating over all backends to find the correct one.
Also remove the now unused to_vnc_backend() helper.
Signed-off-by: Philipp Zabel <philipp.zabel@gmail.com>
Pass the RDP backend to rdp_head_create().
That way the already known backend will not have to be found from the
compositor. This is trivial now, but in the multi-backend case would
entail iterating over all backends to find the correct one.
Also remove the now unused to_rdp_backend() helper.
Signed-off-by: Philipp Zabel <philipp.zabel@gmail.com>
Pass the backend instead of the compositor to the windowed output API
create_head() method and increment the API version.
That way the backend will not have to find the backend pointer from the
compositor. This is trivial now, but in the multi-backend case would
entail iterating over all backends to find the correct one.
Signed-off-by: Philipp Zabel <philipp.zabel@gmail.com>
The output move listener removal was incomplete. Remove the remaining
bits to fix a segfault on shutdown.
Fixes: 40f5eaf401 ("backend-vnc: use output power_state to disable repainting while disconnected")
Signed-off-by: Philipp Zabel <philipp.zabel@gmail.com>
I tried this, and it causes a crash. Leave a note for the future when we
happen to use some other backend with xwayland and find a "leak".
The reason this is a comment and not a Gitlab issue is that you probably
would not go looking for a Gitlab issue saying an idea is a bad one.
This comment is more likely to be found.
It's not really a leak either, it only needs to be fixed if you want a
clean ASan leak report.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.com>
This was not found in the test suite, but if you run wayland-backend
manually with ASan, you see the same leaks as in
backend-headless: fully release pango and fontconfig
Fix them the same way.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.com>
With weston_output_power_on/off() we can use power_state to disable
repainting completely while no VNC client is connected. This allows
to remove the initial repaint and per-output damage tracking.
Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
And use it to exit with using the KEY_ESC, similar to simple-egl.
This was ripped from simple-egl ad litteram.
Signed-off-by: Marius Vlad <marius.vlad@collabora.com>
Rather than having them as an array. This would simplify handling of
maximized and fullscreen and make things easier to reason with.
Signed-off-by: Marius Vlad <marius.vlad@collabora.com>
